Two time scales of adaptation in human learning rates

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry

Background:

• Learning rates, or information updating speeds, vary across situations.
• Adaptation to changing environments can occur via global, meta-learned strategies or local, transient adjustments.

Purpose of the Study:

• To introduce a novel paradigm for measuring and disentangling fast and slow learning rate adaptations.
• To investigate the neural basis of environment-specific learning rate meta-learning.

Main Methods:

• A continuous estimation task (crab fishing) across six distinct environments with varying optimal learning rates.
• Behavioral analysis of learning rate adaptation within and across blocks of trials.
• fMRI data acquisition and representational similarity analysis (RSA).
• Computational modeling to estimate global and transient learning rate components.

Main Results:

• Participants exhibited rapid, within-block learning rate adaptations.
• Evidence for slower, global adaptation of environment-specific learning rates was found, confirmed by initial trials upon re-entering an environment.
• Voxel pattern differences in the central orbitofrontal cortex (OFC) correlated with global environment-specific learning rates.

Conclusions:

• Humans flexibly adjust learning rates on both rapid and gradual timescales.
• The central OFC plays a role in meta-learning by encoding environment-specific information, including optimal learning rates.